Absolute wide-range blood flow measurement method, device and equipment

Abstract

The invention relates to the field of biomedicine, and provides an absolute wide-range blood flow measurement method, device and equipment, and the method comprises the steps: collecting speckle image data containing blood flow information through a laser speckle imaging device; preprocessing the speckle image data to extract time intensity fluctuation frequency characteristics of speckles; the preprocessed speckle image data are input into a trained laser speckle contrast imaging model based on deep learning, and the absolute blood flow velocity is predicted based on the time intensity fluctuation frequency characteristic of speckles; the trained laser speckle contrast imaging model based on deep learning is obtained through training according to a training set, and the training set is preprocessed speckle image data. The problem that in the prior art, the blood flow measurement range is limited is solved, and wide-range absolute blood flow velocity measurement is achieved.

Description

Technical Field

[0001] The present invention relates to the field of biomedical technologies, and in particular, to an absolute wide-range blood flow measurement method, device, and equipment. Background Art

[0002] Currently, the technologies commonly used for blood flow measurement mainly include Transit Time Flow Measurement (TTFM) and Laser Speckle Contrast Imaging (LSCI). As a standard method for cardiac blood flow evaluation, TTFM measures blood flow volume through the difference in ultrasound propagation time. However, in complex surgeries such as Coronary Artery Bypass Grafting (CABG), TTFM can only perform fixed-point measurements on the dissected bypass vessels and is difficult to comprehensively capture the spatio-temporal evolution of blood flow in the distal region of cardiac stenosis.

[0003] On the other hand, LSCI technology infers blood flow characteristics by analyzing the dynamic changes of laser speckles caused by the movement of red blood cells. This technology has advantages such as high spatio-temporal resolution, simple operation, and low cost, and thus has been widely used in blood flow monitoring of superficial tissues (such as the brain, skin, and retina). However, LSCI technology mainly relies on relative blood flow index (rBFI) or speckle contrast (K) as a substitute indicator for blood perfusion. These indicators lack the ability to directly measure absolute blood flow rate and perform poorly in monitoring high blood flow rates (such as myocardial coronary blood flow). In addition, these indicators are also affected by various parameters (such as the proportion of static scattering in the sample, the configuration of experimental equipment, the scattering characteristics of the medium, etc.), resulting in difficulty in obtaining consistent absolute blood flow rates among different samples. Summary of the Invention

[0004] The present invention provides an absolute wide-range blood flow measurement method, device, and equipment, which solves the problem of limited blood flow measurement range in the prior art and realizes wide-range absolute blood flow velocity measurement.

[0005] The present invention provides an absolute wide-range blood flow measurement method, including the following steps: Collect speckle image data containing blood flow information through a laser speckle imaging device; Preprocess the speckle image data to extract the frequency characteristics of the temporal intensity fluctuations of the speckles; Input the preprocessed speckle image data into a trained deep learning-based laser speckle contrast imaging model, and predict the absolute blood flow velocity based on the frequency characteristics of the temporal intensity fluctuations of the speckles; The trained deep learning-based laser speckle contrast imaging model is obtained by training according to a training set, and the training set is the preprocessed speckle image data.

[0006] An absolute wide-range blood flow measurement method provided by the present invention, the laser speckle imaging device includes a near-infrared laser, a camera, a filter, a polarizer, a sleeve lens, and a microscope objective; the camera is used to collect the speckle image of the sample to be measured; the filter, the polarizer, the sleeve lens, and the microscope objective are coaxially arranged in sequence; the filter is used to filter the laser, the polarizer is used to polarize the laser, the sleeve lens and the microscope objective form a microscope system, which is used to magnify the speckle image of the sample to be measured, and the near-infrared laser is used to irradiate the sample to be measured with near-infrared laser, and adjust the light intensity so that the average gray value of the speckle image is 105.

[0007] An absolute wide-range blood flow measurement method provided by the present invention, preprocess the speckle image data, segment the speckle image video into multiple segments containing consecutive frames, and adjust the pixels of the segments to a preset value as the input data of the deep learning-based laser speckle contrast imaging model.

[0008] A trained deep learning-based laser speckle contrast imaging model includes: a plurality of convolution and pooling modules, connected in sequence, used to extract features from the input data; a fully connected layer, connected to the last convolution and pooling module, used to output the probability distribution of blood flow categories; the deep learning-based laser speckle contrast imaging model uses ReLU as the activation function and Adam as the optimizer to accelerate the convergence speed of the model.

[0009] An absolute wide-range blood flow measurement method provided by the present invention, predicting the absolute blood flow velocity based on the time-intensity fluctuation frequency characteristics of the speckle, specifically including: extracting spatio-temporal features in the speckle image layer by layer through a plurality of convolution and pooling modules; after feature extraction, passing the feature map to the fully connected layer, and integrating and transforming the extracted features through the fully connected layer, and finally outputting the probability distribution corresponding to the blood flow velocity category; according to the output probability distribution, determining the predicted absolute blood flow velocity value by looking up the predefined mapping relationship or probability threshold.

[0010] An absolute wide-range blood flow measurement method provided by the present invention, training a deep learning-based laser speckle contrast imaging model according to a training set, specifically including: inputting the preprocessed training set data into the deep learning-based laser speckle contrast imaging model; adjusting the model parameters through the backpropagation algorithm to minimize the loss function of the deep learning-based laser speckle contrast imaging model on the training set; repeating the training process until the model converges or reaches a predetermined number of training epochs, and outputting the trained deep learning-based laser speckle contrast imaging model.

[0011] The present invention also provides an absolute wide - range blood flow measurement device, including the following modules: An acquisition module, configured to acquire speckle image data containing blood flow information through a laser speckle imaging device; A pre - processing module, configured to pre - process the speckle image data to extract the frequency characteristics of the temporal intensity fluctuations of the speckles; A prediction module, configured to input the pre - processed speckle image data into a trained deep - learning - based laser speckle contrast imaging model, and predict the absolute blood flow velocity based on the frequency characteristics of the temporal intensity fluctuations of the speckles; The trained deep - learning - based laser speckle contrast imaging model is obtained by training according to a training set, and the training set is the pre - processed speckle image data.

[0012] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the absolute wide - range blood flow measurement method as described in any one of the above is implemented.

[0013] The present invention also provides a non - transitory computer - readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the absolute wide - range blood flow measurement method as described in any one of the above is implemented.

[0014] The present invention also provides a computer program product, including a computer program. When the computer program is executed by a processor, the absolute wide - range blood flow measurement method as described in any one of the above is implemented.

[0015] The absolute wide - range blood flow measurement method, device, and equipment provided by the present invention have the following beneficial effects: By combining laser speckle imaging and deep - learning technology, using a laser speckle imaging device to capture blood flow information, after pre - processing, the absolute blood flow velocity is predicted through a trained deep - learning model, realizing fast and accurate measurement of blood flow velocity. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the following - described drawings are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0017] Figure 1 It is a flowchart of the absolute wide - range blood flow measurement method provided by the present invention.

[0018] Figure 2 It is a structural schematic diagram of the laser speckle imaging device provided by the present invention.

[0019] Figure 3 is the DL-LSCI model structure provided by the present invention.

[0020] Figure 4 is a schematic diagram of the prediction results of DL-LSCI in in-vivo membrane experiments provided by the present invention.

[0021] Figure 5 is a schematic diagram of the prediction results of DL-LSCI in animal experiments provided by the present invention.

[0022] Figure 6 is a schematic diagram of the structure of the absolute wide-range blood flow measurement device provided by the present invention.

[0023] Figure 7 is a schematic diagram of the structure of the electronic device provided by the present invention.

[0024] Reference numerals: Near-infrared laser 1, camera 2, filter 3, polarizer 4, sleeve lens 5, microscope objective 6, sample to be measured 7. Detailed implementation manners

[0025] To make the objectives, technical solutions and advantages of the present invention clearer, the technical solutions in the present invention will be clearly and completely described below with reference to the accompanying drawings in the present invention. Obviously, the described embodiments are some but not all of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present invention without creative efforts shall fall within the protection scope of the present invention.

[0026] Currently, the techniques commonly used for blood flow measurement are Transit time flow measurement (TTFM) and Laser speckle contrast imaging (LSCI). TTFM is the current standard method for cardiac blood flow evaluation, which measures blood flow through the difference in ultrasound propagation time. However, in coronary artery bypass grafting (CABG), TTFM can only perform fixed-point measurements on the dissected bypass vessels and is difficult to capture the spatio-temporal evolution of blood flow in the distal region of cardiac stenosis, which is crucial for improving the surgical quality control level. The blood flow measurement method based on LSCI infers blood flow characteristics by analyzing the dynamic changes of laser speckles caused by the movement of red blood cells. Due to its advantages such as high spatio-temporal resolution, simple operation, and low cost, it is currently widely used for the monitoring of blood flow in superficial tissues (such as the brain, skin, and retina). This method mainly relies on the relative blood flow index (rBFI) or speckle contrast (K) as a surrogate indicator of blood perfusion, lacks the ability to measure absolute blood flow rate, and is difficult to monitor the blood flow rate of high-flow regions such as myocardial coronary arteries. In addition, these indicators are affected by parameters β , ρ , and n , which are related to the proportion of static scattering in the sample, the configuration of experimental equipment, and the scattering characteristics of the medium, respectively, resulting in difficulty in obtaining consistent absolute blood flow rates among different samples. In recent years, in order to improve the blood flow measurement accuracy of LSCI and obtain absolute blood flow rate, many researchers have tried to combine deep learning techniques to improve the existing methods. Hao et al. used a 3D convolutional neural network to achieve quantitative prediction in the range of 0.08 - 10.74 mm / s, but the estimation range is limited and the performance is poor under static scattering conditions; Chen et al. developed a deep learning method based on UNet and ResNet to reconstruct the 3D blood flow structure in thick tissues, which can perform depth-dependent flow estimation and reduce the bias caused by multiple scattering. However, although these techniques have significantly improved the performance of LSCI, they still cannot provide true absolute blood flow rate measurement in standardized and clinical applications.

[0027] To solve the above problems, the present invention proposes a deep learning-based laser speckle contrast imaging model (DL-LSCI) that predicts absolute blood flow velocity by analyzing the spatio-temporal frequency characteristics of speckle patterns in different blood flows, aiming to obtain a wide range of absolute blood flow rates. The present invention mainly includes two parts, namely a laser speckle imaging device for collecting data and a DL-LSCI model structure for predicting absolute blood flow rates. This method is not only applicable to ideal conditions without static scattering, but also can accurately predict the actual blood flow velocity under static scattering conditions introduced by scattering sheets or biological blood vessels, with a maximum measurable blood flow rate of 308 mm / s. Through experimental verification, this solution shows excellent accuracy and robustness in in-vivo experiments and can provide consistent quantitative results for hemodynamic research under various experimental conditions.

[0028] The following combines Figures 1 - 7 to specifically describe the embodiments of the present invention.

[0029] Figure 1 is a schematic flow chart of the absolute wide-range blood flow measurement method provided by the present invention. As Figure 1 shown, the method includes the following steps: S110. Collect speckle image data containing blood flow information through a laser speckle imaging device.

[0030] According to an absolute wide-range blood flow measurement method provided by the present invention, the laser speckle imaging device includes a near-infrared laser, a camera, a filter, a polarizer, a sleeve lens, and a microscope objective; the camera is used to collect the speckle image of the sample to be measured; the filter, the polarizer, the sleeve lens, and the microscope objective are coaxially arranged in sequence; the filter is used to filter the laser, the polarizer is used to polarize the laser, and the sleeve lens and the microscope objective form a microscope system for magnifying the speckle image of the sample to be measured; the near-infrared laser is used to irradiate the sample to be measured with near-infrared laser and adjust the light intensity so that the average gray value of the speckle image is 105.

[0031] Specifically, as Figure 2 shown is a schematic diagram of the laser speckle imaging device, including a near-infrared laser 1, a camera 2, a filter 3, a polarizer 4, a sleeve lens 5, a microscope objective 6, and a sample to be measured 7.

[0032] The near-infrared laser 1 serves as a light source to generate a stable near-infrared laser beam for imaging illumination.

[0033] The camera 2 is located on the transmission path of the laser beam and is used to capture the laser speckle image scattered back by the sample. The camera 2 is arranged at a certain angle with the laser beam to ensure that the speckle pattern can be clearly received.

[0034] The filter 3 is located in front of the camera and is used to filter out unnecessary spectral components and only allow near-infrared light of a specific wavelength to pass through to enhance image quality.

[0035] The polarizer 4 is located on the optical path in front of the camera to adjust the polarization state of light, reduce the interference of reflected light and scattered light, and improve the signal-to-noise ratio of the image.

[0036] The tube lens 5 and the microscope objective lens 6 are located between the camera 2 and the sample 7 to be tested, and the two form a microscope system for amplifying the speckle image so that the camera can capture a finer speckle structure. The focal length and magnification of the microscope system are adjusted according to specific application requirements.

[0037] The sample 7 to be tested is located at the bottom of the device and is the target object of laser speckle imaging, such as a blood vessel in a biological tissue.

[0038] The above-mentioned components cooperate with each other, and the laser beam generated by the near-infrared laser 1 directly irradiates the sample to be tested 7. The laser scattered back by the sample is amplified by the sleeve lens 5 and the microscope objective 6, and then passes through the polarizer 4 and the filter 3 in sequence, and finally enters the camera 2 for imaging. The various optical elements (filter 3, polarizer 4, sleeve lens 5, microscope objective 6) are arranged in sequence on the transmission path of the laser beam to ensure that the light path is unobstructed, while each plays its specific optical function. The camera 2 is located on one side of the entire device, at a certain angle to the laser beam to capture the scattered laser speckle image. This layout and design enables the laser speckle imaging device to effectively capture and analyze the blood flow information in the sample, providing high-quality image data for subsequent data processing and absolute blood flow velocity measurement.

[0039] In one embodiment of the present invention, the exposure time of the laser speckle imaging device is first unified to 80 μ m, the sampling frequency is 100 Hz, and the blood flow data of 0-462 mm / s are collected by a laser speckle imaging device. Each set of data consists of 200 consecutive frames of speckle patterns. The intensity fluctuations of the speckle patterns at different flow rates are different on the time scale. The lower the blood flow velocity, the more violent the intensity fluctuations.

[0040] S120 , preprocessing the speckle image data to extract the temporal intensity fluctuation frequency characteristics of the speckle.

[0041] According to an absolute wide-range blood flow measurement method provided by the present invention, speckle image data is preprocessed, the speckle image video is segmented into a plurality of segments containing continuous frames, and the pixels of the segments are adjusted to preset values ​​as input data of a laser speckle contrast imaging model based on deep learning.

[0042] Specifically, to ensure the extraction of high signal-to-noise ratio blood flow information from the speckle pattern, during the entire experiment, by adjusting the laser light intensity, the average gray value of the speckle image was stabilized at 105, so that most pixel values remained within the 8-bit dynamic range. Therefore, the present invention can utilize the time-intensity fluctuation frequency characteristics of the speckle as an important basis for estimating blood flow in the DL-LSCI model. In addition, to prevent overfitting of the deep learning model, the present invention expands the dataset through data augmentation methods. The speckle image video is segmented into multiple segments, each segment containing 25 consecutive images, with a step size of 1. For example, the 1st - 25th frames are used as the first segment, the 2nd - 26th frames are used as the second segment, and so on until the last frame. In this way, 176 segments can be obtained for each flow velocity, with a total of 3520 segments, which are divided into a training set and a test set according to a ratio of 7:3. Each segment contains similar speckle change information in the spatio-temporal dimension, while each segment is slightly different. Subsequently, these videos are resized to 112×112 pixels as the input of the DL-LSCI model. Through the above method, the data sufficiency during the model training process and the effectiveness of feature learning are ensured, thereby improving the estimation accuracy of the absolute blood flow rate.

[0043] S130. Input the preprocessed speckle image data into the trained deep learning-based laser speckle contrast imaging model, and predict the absolute blood flow velocity based on the time-intensity fluctuation frequency characteristics of the speckle. The trained deep learning-based laser speckle contrast imaging model is obtained by training according to the training set, and the training set is the preprocessed speckle image data.

[0044] According to an absolute wide-range blood flow measurement method provided by the present invention, the trained deep learning-based laser speckle contrast imaging model includes: a plurality of convolution and pooling modules, connected in sequence, for extracting features from the input data; a fully connected layer, connected to the last convolution and pooling module, for outputting the probability distribution of blood flow categories; the deep learning-based laser speckle contrast imaging model uses ReLU as the activation function and Adam as the optimizer to accelerate the convergence speed of the model.

[0045] According to an absolute wide-range blood flow measurement method provided by the present invention, training the deep learning-based laser speckle contrast imaging model according to the training set specifically includes: inputting the preprocessed training set data into the deep learning-based laser speckle contrast imaging model; adjusting the model parameters through the backpropagation algorithm to minimize the loss function of the deep learning-based laser speckle contrast imaging model on the training set; repeating the training process until the model converges or reaches a predetermined number of training epochs, and outputting the trained deep learning-based laser speckle contrast imaging model.

[0046] An absolute wide-range blood flow measurement method provided by the present invention predicts the absolute blood flow velocity based on the frequency characteristics of the temporal intensity fluctuations of speckles, specifically including: extracting the spatio-temporal features in the speckle image layer by layer through multiple convolution and pooling modules; after feature extraction, passing the feature map to the fully connected layer, integrating and transforming the extracted features through the fully connected layer, and finally outputting the probability distribution corresponding to the blood flow velocity category; according to the output probability distribution, determining the predicted absolute blood flow velocity value by looking up the pre-defined mapping relationship or probability threshold.

[0047] Specifically, as Figure 3 shown, next, a DL-LSCI model is built. After preprocessing such as adjustment and amplification, the data size is 25×112×112, and it passes through five convolution and pooling modules in sequence. One convolution and one pooling operation are performed in the first convolution and pooling module and the second convolution and pooling module. The convolution kernel size is 3×3×3, and the stride is 1×1×1. The pooling layer size in the first convolution and pooling module is 1×2×2, and the pooling layer size in the second convolution and pooling module is 2×2×2. In the third convolution and pooling module, the fourth convolution and pooling module, and the fifth convolution and pooling module, two convolutions and one pooling operation are performed. The convolution kernel size is still 3×3×3, the stride is 1×1×1, and the pooling layer size is 2×2×2. Finally, the size of the obtained feature map is 512×1×4×4. Finally, the feature map passes through two fully connected layers, with 4096 neurons in each layer, and the output is the probability distribution of the blood flow category. In this DL-LSCI model, the activation function uses "ReLU", the optimizer selects Adam to accelerate the convergence speed, the learning rate is 5×10 -5 , and the batch size is 32. The present invention simplifies the blood flow estimation problem into a classification problem, thereby effectively improving the accuracy of absolute blood flow velocity estimation and the convergence performance of the model.

[0048] To verify the effectiveness of the DL-LSCI model in the present invention, in vitro membrane experiments and animal experiments were carried out. In the in vitro membrane experiment, the blood flow velocity was controlled by a stepping motor in the range of 0 mm / s to 462 mm / s, and the corresponding speckle data was collected. Data in four cases were collected: blood flow in a transparent microfluidic chamber, blood flow in a microfluidic chamber covered with a 220-mesh scattering sheet, the coronary artery of a pig, and blood flow in a simulated blood vessel in the internal mammary artery. Taking the above data as the data set, using DL-LSCI to predict its flow velocity, the results are as Figure 4As shown in the figure. The values in the figure represent the predicted probabilities at different flow velocities. The higher the probability value, the more accurate the DL-LSCI prediction. At low flow velocities, the prediction accuracy of the model is above 90%. When the blood flow velocity exceeds 100 mm / s, the probability value drops to 70%-90%, but it can still correctly predict the flow velocity without affecting the overall accuracy. However, when the flow velocity rises to 462 mm / s, the prediction probability drops below 50%, indicating that DL-LSCI cannot effectively estimate the blood flow velocity of 462 mm / s. This is because as the blood flow velocity increases, more speckle spatio-temporal information is integrated within the exposure time (80 μ s), and the dynamic fluctuations become smoother, resulting in a reduced difference between high flow velocities. Therefore, in the models without a covered medium and with a covered diffuser, DL-LSCI can predict blood flow velocities from 0 to 385 mm / s, while for blood samples covered by the other two tissues, this range is from 0 to 308 mm / s. These differences stem from different static scattering components. A lower static scattering component enables the speckle pattern to contain more blood flow information, i.e., a higher signal-to-noise ratio. Therefore, it can be concluded that the DL-LSCI model can achieve an absolute wide range of blood flow measurements from 0 mm / s to 308 mm / s, which is of great significance for significantly improving the accuracy of the LSCI theory. In animal experiments, a rat carotid artery infarction / recovery model was used to obtain different flow velocities, and the blood flow was synchronously recorded using TTFM and the proposed scheme of the present invention. The results are as Figure 5 shown. The abscissa represents the blood flow velocity measured by TTFM, and the ordinate represents the blood flow velocity predicted by DL-LSCI. The smaller dots are the trained samples, and the larger dots are the samples outside the dataset to be evaluated. The dashed line is y = x the reference line for comparing the blood flow velocity measured by TTFM and the blood flow velocity predicted by the model. The right table shows the probability values of blood flow prediction, and all probability values exceed 92%. Therefore, it can be concluded that in animal experiments, the DL-LSCI model can achieve an absolute blood flow measurement from 13 to 105 mm / s.

[0049] The present invention provides a deep learning-based laser speckle contrast imaging model (DL-LSCI) that can estimate a wide range of absolute blood flow rates. By combining three-dimensional convolutional layers with laser speckle characteristics, DL-LSCI realizes the absolute value measurement of blood flow velocity, getting rid of the dependence on parameters of traditional LSCI technology. The model can accurately predict a wide range of blood flow velocities and shows excellent performance in both in vitro membrane experiments and animal experiments. Compared with the clinical experiment gold standard TTFM, the DL-LSCI model improves the blood flow measurement accuracy of laser speckle imaging devices, has the potential for application in high-demand surgeries such as coronary artery bypass grafting (CABG), can achieve regional blood flow imaging visualization, and promotes the development of surgical quality control and biomedical research.

[0050] The absolute wide-range blood flow measurement device provided by the present invention will be described below. The absolute wide-range blood flow measurement device described below can be correspondingly referred to the absolute wide-range blood flow measurement method described above.

[0051] As Figure 6 shown, an absolute wide-range blood flow measurement device provided by the present invention includes: An acquisition module 610, configured to acquire speckle image data containing blood flow information through a laser speckle imaging device; A preprocessing module 620, configured to preprocess the speckle image data to extract the time-intensity fluctuation frequency characteristics of the speckles; A prediction module 630, configured to input the preprocessed speckle image data into a trained deep learning-based laser speckle contrast imaging model, and predict the absolute blood flow velocity based on the time-intensity fluctuation frequency characteristics of the speckles; The trained deep learning-based laser speckle contrast imaging model is obtained by training according to a training set, and the training set is the preprocessed speckle image data.

[0052] Figure 7 Illustrated is a schematic physical structure diagram of an electronic device. As Figure 7 shown, the electronic device may include: a processor 710, a communication interface 720, a memory 730, and a communication bus 740. Among them, the processor 710, the communication interface 720, and the memory 730 complete communication with each other through the communication bus 740. The processor 710 may call logical instructions in the memory 730 to execute the absolute wide-range blood flow measurement method, and the method includes: acquiring speckle image data containing blood flow information through a laser speckle imaging device; preprocessing the speckle image data to extract the time-intensity fluctuation frequency characteristics of the speckles; inputting the preprocessed speckle image data into a trained deep learning-based laser speckle contrast imaging model, and predicting the absolute blood flow velocity based on the time-intensity fluctuation frequency characteristics of the speckles; the trained deep learning-based laser speckle contrast imaging model is obtained by training according to a training set, and the training set is the preprocessed speckle image data.

[0053] In addition, when the logical instructions in the above-mentioned memory 730 are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods in various embodiments of the present invention. The aforementioned storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs that can store program codes.

[0054] On the other hand, the present invention also provides a computer program product. The computer program product includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the absolute wide-range blood flow measurement method provided by the above-mentioned various methods. The method includes: collecting speckle image data containing blood flow information through a laser speckle imaging device; preprocessing the speckle image data to extract the time-intensity fluctuation frequency characteristics of the speckles; inputting the preprocessed speckle image data into a trained deep learning-based laser speckle contrast imaging model, and predicting the absolute blood flow velocity based on the time-intensity fluctuation frequency characteristics of the speckles; the trained deep learning-based laser speckle contrast imaging model is obtained by training according to a training set, and the training set is the preprocessed speckle image data.

[0055] On another aspect, the present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it is implemented to execute the absolute wide-range blood flow measurement method provided by the above-mentioned various methods. The method includes: collecting speckle image data containing blood flow information through a laser speckle imaging device; preprocessing the speckle image data to extract the time-intensity fluctuation frequency characteristics of the speckles; inputting the preprocessed speckle image data into a trained deep learning-based laser speckle contrast imaging model, and predicting the absolute blood flow velocity based on the time-intensity fluctuation frequency characteristics of the speckles; the trained deep learning-based laser speckle contrast imaging model is obtained by training according to a training set, and the training set is the preprocessed speckle image data.

[0056] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art can understand and implement it without creative work.

[0057] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on such an understanding, the essence of the above technical solution, or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods of each embodiment or some parts of the embodiments.

[0058] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments or equivalently replace some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for measuring blood flow in an absolute wide range, characterized in that: include: Acquiring speckle image data containing blood flow information by means of a laser speckle imaging device; Preprocessing the speckle image data to extract the temporal intensity fluctuation frequency characteristics of the speckle; Inputting the preprocessed speckle image data into a trained laser speckle contrast imaging model based on deep learning, and predicting the absolute blood flow velocity based on the temporal intensity fluctuation frequency characteristics of the speckle; The trained deep learning-based laser speckle contrast imaging model is obtained by training according to a training set, and the training set is preprocessed speckle image data.

2. The absolute wide range blood flow measurement method according to claim 1, characterized in that: The laser speckle imaging device comprises a near-infrared laser, a camera, a filter, a polarizing plate, a tube lens and a microscope objective lens; The camera is used to collect the speckle image of the sample to be tested; The filter, polarizing plate, sleeve lens and microscope objective lens are coaxially arranged in sequence; The filter is used to filter the laser, the polarizer is used to polarize the laser, and the tube lens and the microscope objective lens form a microscope system for amplifying the speckle image of the sample to be tested; The near-infrared laser is used to irradiate the sample to be tested with the near-infrared laser, and the light intensity is adjusted so that the average gray value of the speckle image is 105.

3. The absolute wide range blood flow measurement method according to claim 1, characterized in that: The speckle image data is preprocessed, the speckle image video is segmented into a plurality of segments including continuous frames, and the pixels of the segments are adjusted to preset values ​​as input data of a laser speckle contrast imaging model based on deep learning.

4. The absolute wide range blood flow measurement method according to claim 1, characterized in that: The trained deep learning-based laser speckle contrast imaging model includes: Multiple convolution and pooling modules are connected in sequence to extract features from input data; The fully connected layer is connected to the last convolution and pooling module to output the probability distribution of blood flow categories; The laser speckle contrast imaging model based on deep learning adopts ReLU as the activation function and Adam as the optimizer to accelerate the convergence speed of the model.

5. The absolute wide range blood flow measurement method according to claim 4, characterized in that: The predicting of the absolute blood flow velocity based on the temporal intensity fluctuation frequency characteristics of the speckle specifically includes: The spatiotemporal features in the speckle image are extracted layer by layer through multiple convolution and pooling modules; After feature extraction, the feature map is passed to the fully connected layer, through which the extracted features are integrated and transformed, and the final output is the probability distribution of the corresponding blood flow velocity category; According to the output probability distribution, the predicted absolute blood flow velocity value is determined by searching a predefined mapping relationship or probability threshold.

6. The absolute wide range blood flow measurement method according to claim 1, characterized in that: The laser speckle contrast imaging model based on deep learning is trained according to the training set, including: The preprocessed training set data is input into the laser speckle contrast imaging model based on deep learning; Adjusting the model parameters by a back-propagation algorithm to minimize the loss function of the deep learning-based laser speckle contrast imaging model on the training set; The training process is repeated until the model converges or reaches a predetermined number of training rounds, and the trained laser speckle contrast imaging model based on deep learning is output.

7. An absolute wide range blood flow measurement device, characterized in that: include: An acquisition module, used for acquiring speckle image data containing blood flow information through a laser speckle imaging device; A preprocessing module, used for preprocessing the speckle image data to extract the temporal intensity fluctuation frequency characteristics of the speckle; A prediction module, used for inputting the preprocessed speckle image data into a trained laser speckle contrast imaging model based on deep learning, and predicting the absolute blood flow velocity based on the temporal intensity fluctuation frequency characteristics of the speckle; The trained deep learning-based laser speckle contrast imaging model is obtained by training according to a training set, and the training set is preprocessed speckle image data.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the absolute wide-range blood flow measurement method according to any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the absolute wide-range blood flow measurement method according to any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the absolute wide-range blood flow measurement method according to any one of claims 1 to 6 is implemented.

Images